RNA-binding proteins regulate aldosterone homeostasis in ...

RNA-binding proteins regulate aldosterone homeostasisin human steroidogenic cells

RUI FU,1,4 KIMBERLY WELLMAN,1,4 AMBER BALDWIN,1 JUILEE REGE,2 KATHRYN WALTERS,1

ANTJE HIRSEKORN,3 KENT RIEMONDY,1 WILLIAM E. RAINEY,2 and NEELANJAN MUKHERJEE1

1RNA Bioscience Initiative, University of Colorado School of Medicine, Aurora, Colorado 80045, USA2University of Michigan School of Medicine, Ann Arbor, Michigan 48109, USA3Max Delbrück Center for Molecular Medicine, 13092 Berlin, Germany

ABSTRACT

Angiotensin II (AngII) stimulates adrenocortical cells to produce aldosterone, amaster regulator of blood pressure. Despiteextensive characterization of the transcriptional and enzymatic control of adrenocortical steroidogenesis, there are stillmajor gaps in the precise regulation of AngII-induced gene expression kinetics. Specifically, we do not know the regulatorycontribution of RNA-binding proteins (RBPs) and RNA decay, which can control the timing of stimulus-induced gene ex-pression. To investigate this question, we performed a high-resolution RNA-seq time course of the AngII stimulation re-sponse and 4-thiouridine pulse labeling in a steroidogenic human cell line (H295R). We identified twelve temporallydistinct gene expression responses that contained mRNA encoding proteins known to be important for various steps ofaldosterone production, such as cAMP signaling components and steroidogenic enzymes. AngII response kinetics formany of these mRNAs revealed a coordinated increase in both synthesis and decay. These findings were validated in pri-mary human adrenocortical cells stimulated ex vivo with AngII. Using a candidate screen, we identified a subset of RNA-binding protein and RNA decay factors that activate or repress AngII-stimulated aldosterone production. Among the re-pressors of aldosterone were BTG2, which promotes deadenylation and global RNA decay. BTG2was induced in responseto AngII stimulation and promoted the repression of mRNAs encoding prosteroidogenic factors indicating the existence ofan incoherent feedforward loop controlling aldosterone homeostasis. These data support a model in which coordinatedincreases in transcription and decay facilitate the major transcriptomic changes required to implement a prosteroidogenicexpression program that actively resolves to prevent aldosterone overproduction.

Keywords: RNA decay; RNA-binding proteins; gene expression kinetics; transcriptomics; steroid hormone

INTRODUCTION

Angiotensin II (AngII) is a potent octapeptide hormone thatbinds to angiotensin II receptor type 1 (AGTR1) in adrenalzona glomerulosa cells to stimulate the production of themineralocorticoid aldosterone from cholesterol. The pri-mary function of aldosterone is to regulate blood pressureand volume through the control of renal salt excretion(Gross and Gysel 1954). Additional functions of aldoste-rone discovered more recently include directly mediatingcardiac and vascular remodeling (Feldman 2014;Monticone et al. 2018). Aldosterone is a small, nonpolar,and hydrophobic steroid hormone that passes throughthe cell membrane upon production. Given the lack of in-tracellular aldosterone, it must be produced de novo in re-

sponse to steroidogenic stimuli. Thus, it is critical to tightlycontrol the timing and amount of de novo aldosterone pro-duction. Pathological aldosterone excess, such as in prima-ry aldosteronism, leads to hypertension and increasedcardiovascular risk in humans (Monticone et al. 2018;Nanba et al. 2019; Brown et al. 2020). Although the stimu-latory effect of AngII upon aldosterone secretion has beenextensively studied, the temporal regulation of the AngIIresponse, which is critical for maintaining aldosterone ho-meostasis, remains poorly understood.The temporal coordination of AngII-stimulated aldoste-

rone biosynthesis is described in two stages: an “early reg-ulatory step” (minutes) and a “late regulatory step” (hours)(Nogueira et al. 2009; Hattangady et al. 2012). The earlyregulatory step is characterized by rapid signaling path-ways promoting cholesterol mobilization. A key part of

4These authors contributed equally to this work.Corresponding author: [email protected] is online at http://www.rnajournal.org/cgi/doi/10.1261/rna.

078727.121. Freely available online through the RNA Open Accessoption.

© 2021 Fu et al. This article, published in RNA, is available under aCreative Commons License (Attribution 4.0 International), as describedat http://creativecommons.org/licenses/by/4.0/.

RNA (2021) 27:933–945; Published by Cold Spring Harbor Laboratory Press for the RNA Society 933

the early step is inducing the expression of SteroidogenicAcute Regulatory Protein (STAR) that transports cholester-ol from the outer to inner mitochondrial membrane allow-ing for conversion by CYP11A1 into the pregnenolone,which is a precursor for the production of all steroid hor-mones including aldosterone. The late regulatory step ischaracterized by inducing the expression of steroid biosyn-thesis enzymes such as aldosterone synthase (CYP11B2).Although the temporal order of the AngII gene expressionresponse underlies both regulatory steps, there has notbeen an examination of steroidogenic transcriptome dy-namics with sufficient temporal resolution to systematicallyidentify distinct response kinetics.

There are two prototypical examples of coordinatedtemporal expression responses that feature different regu-latory strategies. One example utilizes a cascade of tran-scription factor activity to generate waves of temporalexpression patterns. Transcriptional cascades have beenobserved in the yeast cell cycle and innate immune re-sponses (Luscombe et al. 2004; Smale 2012). The other ex-ample relies on intrinsic differences in the stability betweeninduced transcripts to generate sequential response pat-terns. Specifically, the time to peak response is inverselycorrelated with the stability of the transcript, and thus un-stable RNAs will exhibit more rapid induction and returnto baseline than stable RNAs (Palumbo et al. 2015). Thispost-transcriptionally regulatedmodel of temporal coordi-nation of response kinetics has been observed largely inimmune and inflammation responses (Hao and Baltimore2009; Elkonet al. 2010; Rabani et al. 2011). Interestingly, in-creasing the decay rate of transcriptionally induced RNAsresults in a kinetic response with a quicker time to peak ex-pression and return to baseline. This strategy of increasingsynthesis and decay has been observed in the yeast oxida-tive stress response and immune activation of mouse den-dritic cells (Shalem et al. 2008; Rabani et al. 2014). A lack ofresponse resolution in both of those biological systemswould lead to deleterious effects for the cell and organism,much like the uncontrolled resolution of aldosteroneproduction.

RNA-binding proteins (RBPs) are a class of trans-actingregulatory factors that control all aspects of RNA metabo-lism (Gerstberger et al. 2014).Many RBPs havebeen shownto control the stability of mRNAs through interactions withcis-regulatory elements in the 3′-UTR of their target RNAs(Keene 2007). RBPs binding to AU-rich elements are oneof the most well-characterized examples for the regulationof RNA stability and have been associated with controllingthe temporal order and shape of immune response kinetics(Hao and Baltimore 2009; Elkon et al. 2010; Rabani et al.2014). General regulators of RNA stability, such as factorsthat modify deadenylation, can also control the temporalresponse kinetics of induced RNAs and may do so in cardi-ac hypertrophy (Mauxion et al. 2008;Masumura et al. 2016;Stupfler et al. 2016). In spite of the importnace of RBPs in

the regulation of decay, expression kinetics, andmany oth-er post-transcriptional processes, their role in AngII-medi-ated aldosterone production is poorly characterized.

To understand the molecular basis and role of RNA reg-ulation on the kinetics of the AngII-aldosterone gene reg-ulatory program, we measured transcriptome dynamics athigh-resolution applying RNA-seq on a steroidogenic hu-man cell line (H295R) stimulated with AngII. While tran-scriptional control was clearly prominent, RNA decayrates determined the time to maximal change in expres-sion. Furthermore, we observed AngII-dependent increas-es in RNA decay for many transcriptionally induced RNAs,including those encoding key prosteroidogenic proteins.These findings were validated in primary human adreno-cortical cells stimulated ex vivo. We identified RBPs andRNA decay factors that either activate or repress aldoste-rone production. Finally, we found that AngII-inducedRNA decay factor BTG2 represses aldosterone pro-duction and promotes the clearance of steroidogenicRNAs. Together, these data support a model in whichcoordinated increases in transcription and RNA decay fa-cilitates the major transcriptomic remodeling required toimplement a prosteroidogenic gene expression programthat is temporally restricted to prevent aldosteroneoverproduction.

RESULTS

Identification of genes with temporally distinctAngII-response kinetics in H295R cells

To delineate the temporal coordination of the AngII-in-duced gene expression response, we performed a high-resolution (twelve time points in duplicate) AngII stimula-tion RNA-seq time course in H295R cells (Fig. 1A). Ratherthan poly(A)-enrichment, we performed rRNA depletionto enable measurement of both precursor and matureRNAs. The data were high quality as samples clustered firstby replicates (r∼0.99) and then time post-stimulation(Supplemental Fig. 1A,B). First, we focused on changes inmature RNA (sum of all mature transcripts per gene,Supplemental Table 1). PCA analysis clearly indicated dif-ferences in gene expression as early as 15minwith the larg-est remodeling of the transcriptomeoccurring∼6–8hpost-stimulation (Fig. 1B). Furthermore, the return toward theunstimulated state by 24 h suggests that this time coursecaptures most of the AngII-induced changes in RNA levelsin this system. Next, we determined the genes exhibitingstatistically significant changes (FDR≤0.001 and greaterthan twofold change) in mature RNA expression at anytime point of AngII treatment versus unstimulated (n=2417). The changes in expression measured by RNA-seqwere recapitulated (R>0.95) by qRT-PCR for a subset ofdifferentially expressed genes (Supplemental Fig. 1C).Differentially expressed coding and long-noncoding

Fu et al.

934 RNA (2021) Vol. 27, No. 8

RNAs both exhibited the largest change in expression be-tween 3 and 8 h (Fig. 1C). Intriguingly, snRNAs appeared tobe up-regulated rapidly, peaking around 2 h before return-ing to baseline.

To identify genes with similar temporal changes in ex-pression, we performed K-means clustering on the matureexpression level changes for the differentially expressedgenes. This revealed 12 groups of genes with temporally

E

F

B

A

C

D

FIGURE 1. High-resolution AngII-response kinetics in H295R cells. (A) Cartoon of AngII treatment time course RNA-sequencing experiment.Total RNA is depleted for rRNA and precursor and mature RNA expression levels are calculated allowing modeling of synthesis and decay. (B)Scatter plot of the first two principal components calculated on a matrix of the mean expression level for each expressed gene at each time point(colors) after AngII stimulation. (C ) Violin plot of the distribution of genes exhibiting statistically significant AngII-induced changes in expressionper timepoint and category. (D) Line plot depicting the average change in expression (error bar represents standard error) relative to unstimulatedcells for each group of genes with similar response kinetics determined by k-means clustering. Each color represents a specific group of geneswith the name defined by the direction, shape, and timing of the AngII-induced response. (E) Heatmap of selected gene sets (Molecular Functionontology) that are significantly enriched in the AngII response groups. (F ) Heatmap of log2-based fold changes in pre-mRNA and mature mRNAlevels upon AngII stimulation compared to unstimulated cells for mRNAs encoding key steroidogenic enzymes.

RNA regulatory control of aldosterone production

www.rnajournal.org 935

distinct AngII-response kinetics (Fig.1D; Supplemental Fig. 1D; Sup-plemental Table 2). The groups werelabeled based on the direction (acti-vated/repressed), magnitude (strong/weak), and timing of expressionchanges. For example, RNAs belong-ing to as1 (activated strong 1), exhibit-ed an average maximal activation of∼20-fold peaking around 1 to 2 h,while RNAs in as4 (activated strong 4)exhibited an average maximal activa-tion of approximately fourfold peakingaround 7 h. The as1 RNAs encodedproteins involved in the calcium andcyclic AMP signaling pathway that areactivated during steroidogenesis (Fig.1E; Supplemental Fig. 1E), includingNR4A1 and NR4A2, which are knowntranscriptional regulators of enzymescontrolling aldosterone production(Bassett et al. 2004; Romero et al.2004). Unexpectedly, we found a verystrong enrichment for snRNAs, andparticularly U1 snRNA, in aw1 (Supple-mental Fig. 1F). We observed a strongenrichment in as2 for mRNAs encod-ing proteins regulating the unfoldedprotein response. RNA-binding pro-teins that control transcript-specificRNA decay were enriched in as2 and aw2. The aw3 andas3 groups were enriched for mRNAs encoding steroid bio-synthesis proteins. The induction of mRNAs encoding ste-roidogenic enzymes required for the conversion ofcholesterol to aldosterone was consistent with previousstudies validating the quality of the data (Fig. 1F; Romeroet al. 2007;Wang et al. 2012). Altogether, our analysis dem-onstrated that AngII response kinetics exhibited temporalcoordination of functionally related genes.

RNA decay controls AngII response kinetics

Simulations and previous studies in other stimulation-re-sponse paradigms have shown that the initial RNA decayrates in the unstimulated cells govern the time to peak ex-pression response and thus the shape of the expression re-sponse (Rabani et al. 2011; Palumbo et al. 2015).Importantly, this phenomenon assumes constant decayrates throughout the response. As a starting point, wequantified the time to maximal expression change foreach gene (see Materials and Methods), which increasedconcordantly with peak time assigned to each group(Fig. 2A, left). For example, genes in as1 (median 1.8 h)reached their peak expression 5.7× faster than those inas4 (median 10.9 h). Next to test if RNA decay rates gov-

erned maximal response kinetics in our system, we mea-sured decay rates in unstimulated H295R cells using RNAmetabolic labeling as we and others have done previously(Milek et al. 2017; Mukherjee et al. 2017). Cells werepulsed with 4-thiouridine (4sU) for 20 min and both totalRNA and 4sU-labeled RNAwere subject to rRNAdepletionand RNA-sequencing, and synthesis, processing, and de-cay rates were calculated using INSPEcT for 11,668 genesin unstimulated H295R cells (Supplemental Fig. 2A–C;Supplemental Table 3; de Pretis et al. 2015). Our decayrates were consistent with established expectations for sta-ble and unstable mRNAs confirming the validity of ourmeasurements. As expected, transcripts encoding ribo-somal proteins had low decay rates, while transcripts en-coding immediate early genes had high decay rates(Supplemental Fig. 2D). We found higher decay rates forRNAs in clusters that peaked earlier post AngII stimulation(Fig. 2B, right). For example, RNAs in as1 had a ∼4.5×higher decay rate than RNAs in as4. Indeed, we observedan inverse relationship between the median decay rateand the median time to maximal induction for bothstrongly and weakly activated genes (Fig. 2B). These re-sults demonstrated that initial RNA decay rates in unstimu-lated cells play an important role in shaping AngII-response kinetics.

BA

FIGURE 2. The influence of RNA decay on AngII response kinetics. (A) Distribution of time topeak response for all RNAs per cluster (left). Distribution of decay rates in unstimulated cells forall RNAs per cluster determined from 4sU-labeling experiments (right). (B) Scatter plot of theaverage decay rate and peakiness for activated RNAs. (C ) Scatter plot of the P-value and en-richment of RBPmotifs from in the top 500 7mers found in the 3′-UTR of RNAs with high decayrates in unstimulated cells.

Fu et al.

936 RNA (2021) Vol. 27, No. 8

Next, we ascertained putative regulatory mechanismscontrolling RNA decay rates in steroidogenic cells. Toidentify possible RBPs or miRNAs regulating stability, werankedmRNAs by decay rates and used cWords to identify7-mers overrepresented in the 3′-UTR sequence of unsta-ble and stable RNAs (Rasmussen et al. 2013). Most of thesignificantly enrichedmotifs were associated with unstableRNAs versus stable RNAs. Therefore, we scored the top500 7mers associated with instability for matches to RBPrecognition motifs from external databases. These analy-ses revealed a striking enrichment for RBPs binding AU-rich elements (AREs), which are key determinants of RNAstability (Fig. 2C). We also observed enrichment for RBPsregulating splicing and export, such as U2AF2 and RALY.These data identify putative RBPs and RNA regulatory ele-ments that regulate RNA stability in unstimulated H295Rcells. Assuming constant RNA decay rates throughoutthe AngII-response, higher decay rates in the unstimulatedcells inversely correlate with time to peak expression in re-sponse to AngII stimulation (Fig. 2B). Therefore, RBPs thatcontrol the RNA decay rates in unstimulated cells are im-portant regulatory factors controlling AngII response kinet-ics and ultimately steroidogenesis.

Increased RNA decay during steroidogenesis

Having established the importance of initial RNA decayrates on AngII response kinetics, we next asked if therewasevidence for changes indecay rateduring the steroido-genic response. We compared changes in pre-mRNA ver-sus mRNA levels to determine whether gene expressionchanges were better explained by transcriptional or post-transcriptional regulation. Specifically, we utilized readsaligning to introns versus exons to determine AngII-in-duced changes in levels of pre-mRNA and mRNA, respec-tively (Gaidatzis et al. 2015; Alkallas et al. 2017; Mukherjeeet al. 2017, 2019). We added the entire precursor RNA se-quence and treated it as an additional transcript isoform tospecifically quantify precursor (Supplemental Table 4) andmature transcript levels using Salmon (Patro et al. 2017). Asa first pass, we calculated the correlation between thechanges in pre-mRNAandmRNA levels of genes encodingsteroidogenic enzymes required for the conversion of cho-lesterol to aldosterone and found examples that are largelyexplained by changes in transcriptional regulation (highcorrelation), as well as examples indicative of changes inpost-transcriptional regulation (low correlation) (Fig. 3A).The same analysis was performed on all ∼1500 differen-tially expressed genes with introns and sufficient intronicread support and revealed a broad distribution of correla-tion coefficients (Fig. 3B). Each temporal group had manygenes with low or negative correlation, indicating thatchanges in transcriptional regulation alone cannot explainthe mature RNA changes. We also performed cross-corre-lation to account for processing delays and observed simi-

lar results (Supplemental Fig. 3A). These results suggestedthat many genes were experiencing stabilization or desta-bilization in response to AngII stimulation.To test the extent and direction of changes in posttran-

scriptional regulation, we performed exon-intron-splitanalysis (EISA) using precursor and mature RNA estimatescomparing all AngII-stimulated samples to the unstimu-lated baseline (see Materials and Methods; SupplementalTable 5; Gaidatzis et al. 2015). Filtering of the initial 2417differentially expressed genes for the presence of intronsand sufficient intronic read coverage resulted in 1194genes that were analyzed. Among these genes a largemajority (84% n=223) showed evidence of mRNA destabi-lization. Destabilized transcripts were significantly overrep-resented in activated genes that had peak expressionbetween 4–8 h post AngII stimulation (as3, as4, aw2,aw3) (Fig. 3C). Next, we examined AngII-induced geneswith evidence for destabilization focusing on as4 andaw3, which contained many steroidogenic genes (Fig.1E). The amplitude of AngII-induced changes in precursorRNAwere similar between genes with (blue lines) andwith-out evidence for destabilization (black lines), albeit the pre-cursor RNAs for destabilized genes tend to peak earlier(Fig. 3D, left). However, the changes in the mature RNAof destabilized genes (blue lines) were clearly suppressedrelative to genes without evidence for post-transcriptionalregulation (black lines) (Fig. 3D, right). Included among thedestabilized mRNAs were proteins encoding STAR, whichis required for the proper and timely enzymatic conversionof cholesterol to steroid hormones, and MC2R (Lin et al.1995), which is the ACTH receptor known to be inducedby AngII stimulation (Fig. 3E; Lebrethon et al. 1994; Parmaret al. 2008). We identified potential RBPs responsible forthe increased decay by searching for RBP binding motifsoverrepresented in the 3′-UTR of destabilized mRNAs be-longing to aw3 and other clusters (Fig. 3F; SupplementalFig. 3B) enriched for destabilized mRNAs. Among theseRBPs are known regulators of RNA decay ILF2, HNRNPL,CNOT4 of the CCR4-NOT deadenylation complex (Albertet al. 2000; Shim et al. 2002; Hui et al. 2003). None of thecandidates exhibited differential expression in response toAngII, nevertheless their activity or protein levels couldchange in the absence of changes in mRNA levels. Alto-gether, we have identified hundreds of genes (n=223) ex-hibiting coordinate increases in both transcription andRNA decay, including key steroidogenic genes, and puta-tive RBPs regulating AngII-induced post-transcriptionalregulatory dynamics.

Ex vivo steroidogenic response in primary humanadrenocortical cells

Although the expression dynamics in H295R cells stimulat-ed with AngII suggested a key role for RBP-mediated RNAdecay, the extent to whether these expression changes

RNA regulatory control of aldosterone production

www.rnajournal.org 937

occur in normal human adrenocortical steroidogenesis isunclear. Therefore, we performed RNA-seq in adrenocor-tical cells isolated from the adrenal gland of a tissue donorand treated ex vivo with either AngII or ACTH or basal me-dia for 3 or 24 h in triplicate (Fig. 4A; Supplemental Tables6, 7; Xing et al. 2010, 2011). We included ACTH, whichstimulates the production of cortisol by the zona fascicu-lata, because H295R cells do not correspond to a specificadrenocortical zone but produce cortisol and androgensin addition to aldosterone when stimulated with AngII(Parmar et al. 2008). Replicates were highly correlatedand clustered by treatment condition and time (Supple-mental Fig. 4A). PCA analysis revealed that compared tobasal media cells treated with ACTH exhibited more dif-ferences than those treated with AngII (Fig. 4B). Acrossall conditions, we detected 3217 genes with significantlydifferent expression levels (FDR<0.05). Consistently, wedetected substantially more statistically significant expres-sion changes induced by ACTH versus AngII at 3 h and 24h (Supplemental Fig. 4B, left and right, respectively). How-

ever, the expression changes for genes with significantdifferential expression upon either ACTH or AngII treat-ment (n=3217) were strongly positively correlated. Thisindicated that the differences between ACTH and AngIIwere more quantitative (magnitude of the expressionchange) rather than qualitative (completely differentgenes changing) (Fig. 4C). Consistent with hormone pro-duction, both AngII and ACTH resulted in the inductionof genes encoding key steroidogenic enzymes (Fig. 4D).Finally we tested if the AngII-induced changes were reca-pitulated in the ex vivo stimulation paradigm using geneset enrichment analysis (GSEA) (Tamayo et al. 2005). In-deed, as2 genes, which were robustly and rapidly inducedby AngII in H295R cells, were significantly up-regulated af-ter ex vivo treatment of primary adrenocortical cells withACTH (Fig. 4E, top left) and AngII (Fig. 4E, bottom left)for 3 h. Likewise, as4 genes, which also exhibited a robustalbeit delayed induction in response to AngII in H295Rcells, were significantly up-regulated after ex vivo treat-ment of primary adrenocortical cells with ACTH (Fig. 4E,

EF

BA C D

FIGURE 3. Dynamic changes in RNA decay during steroidogenesis. (A) Line plot of fold change in pre-mRNA (black) and mRNA (red) expressionfor mRNAs encoding enzymes responsible for conversion of cholesterol to aldosterone (top-to-bottom). Pearson correlation coefficient betweenchanges pre-mRNA andmRNA for each gene (top right). (B) Distribution of Pearson correlation coefficient between pre-mRNA andmRNA chang-es for each gene across all time points for each response cluster. (C ) Heatmap of the odds ratio for the overlap between response cluster mem-bership and RNAs exhibiting evidence for stabilization, destabilization, or neither as determined by EISA (yellow indicates more overlap). (D) Lineplot of median fold change in pre-mRNA (left) and mRNA (right) expression for genes in as4 (top) or aw3 (below) with either evidence for desta-bilization (blue) or no change in posttranscriptional regulation (black). Error bar represents standard error. (E) Line plot of mean fold change in pre-mRNA (black) andmRNA (red) expression for example genes from as3 (top) or aw3 (below) and evidence for destabilization (left) or no difference indecay rates (right). (F ) Scatter plot of the P-value and enrichment of RBPmotifs in 5mers within the 3′-UTR of destabilizedmRNAs in aw3 versus 3′-UTRs of nondifferentially expressed mRNAs.

Fu et al.

938 RNA (2021) Vol. 27, No. 8

top right) and AngII (Fig. 4E, bottom right) for 24 h. Wefound 154 RBPs among the union of differential genes, in-cluding ones that were differentially expressed upon AngIItreatment of H295R cells, such as CPEB4, MBNL1,

MBNL2, MSI2, PEG10, ZFP36, and ZFP36L2. Overall, wefound strong concordance between the direction and tim-ing of expression changes in H295R cells and primary cells(Supplemental Fig. 4C).

E

B

A

C D

FIGURE 4. Gene expression dynamics of ex vivo stimulated primary human adrenocortical cells. (A) Ex vivo stimulation of cortical cells isolatedfrom normal human adrenal glands treated with AngII, ACTH, or basal media for 3 or 24 h. Strand-specific paired-end RNA-seq was performed intriplicate on polyadenylated RNA. (B) Scatter plot of the first two principal components calculated on a matrix of the mean expression level foreach expressed gene at each time (label) and treatment (color). (C ) Bar plot of the Pearson correlation coefficient between AngII-induced andACTH-induced changes in gene expression versus basal cells for the union of genes exhibiting statistically significant changes in expression inany condition/time. Correlation coefficients for comparisons of the same stimulation time are colored red. (D) Heatmap of log2 fold change inexpression difference for stimulated (ACTHon left, AngII on right) versus basal for mRNAs encoding key steroidogenic enzymes. (E ) GSEA runningenrichment plots depicting the enrichment of as2 (left) and as4 (right) gene sets representing clusters of geneswith similar AngII-induction kineticsin H295R cells (see Fig. 1D) in a list of genes ranked by the fold change upon ex vivo stimulation of primary human adrenocortical cells with ACTH(top) or AngII (basal) versus basal media.

RNA regulatory control of aldosterone production

www.rnajournal.org 939

RNA-binding proteins regulate aldosterone levels

Given the agreement in direction and timing of the steroi-dogenic gene expression response between H295R andprimary human cells, we performed an siRNA screen toidentify RBPs that regulate aldosterone levels in H295Rcells. We selected 16 candidates from an annotated listof 1542 human RBPs (see Gerstberger et al. 2014) basedon (i) differential expression upon AngII treatment inH295R cells; (ii) high expression in H295R cells; (iii) highand adrenal gland-specific RNA expression across normalhuman tissues; and (iv) RBP motifs enriched in the 3′-UTRsof unstable and destabilized mRNAs (Fig. 5A). Due to ourinclusion criteria the candidate RBPs were not all knowndecay factors (e.g., TRUB1 or PELO). We used indepen-dent siRNAs for each candidate RBP and measured aldo-sterone levels in cell supernatants 24 h after treatmentwith either vehicle or AngII (see Materials and Methodsfor details). Aldosterone levels were normalized by cell vi-ability and compared to a mock electroporation control forboth unstimulated and stimulated cell supernatants(Supplemental Fig. 5A; Supplemental Table 8). As expect-ed, the depletion of the key steroidogenic transcriptionfactor SF-1 (NR5A1) resulted in loss of aldosterone. FiveRBPs exhibited statistically significant increases in aldoste-rone levels by two independent siRNAs in AngII stimulatedcell supernatants (Fig. 5B). Among these putative repres-sors of aldosterone were regulators of global RNA decay(BTG2), ARE-decay (ZFP36L1 and ZFP36L2), and a pseu-douridine synthase (TRUB1). We identified three potentialRBPs for which two independent siRNA knockdowns re-sulted in decreased aldosterone levels suggesting theypromote aldosterone production (Fig. 5C). Specifically,these putative activators of aldosterone were a splicingfactor (MBNL2), an RNA decay factor (CPEB4), and a sig-naling scaffold and RNA localization factor (AKAP1). Forthe seven other RBPs we either observed no changeand/or conflicting results for the different siRNAs(HNRNPL, MBNL1, PEG10, PNRC2, ZFP36), or we onlyscreened a single siRNA (MSI2 and PELO) (Fig. 5D).Knockdown of RBPs resulted in similar, but more subtle ef-fects on aldosterone levels in unstimulated cells, which wasto be expected given the minimal amounts of aldosteroneproduced in the absence of AngII (Supplemental Fig. 5B–D). Interestingly we observed potential functional redun-dancy between the ZFP36 family members. DepletionZFP36L2, which is the most abundant family memberby mRNA expression, had the most consistent and potentincrease in aldosterone levels (Supplemental Fig. 5E).These data revealed eight RBPs (10 including RBPs target-ed by a single siRNA) that either activated or repressed al-dosterone production. Seven of the ten RBPs wereinduced by AngII stimulation, which could indicate an in-crease in their respective post-transcriptional regulatoryactivity (Supplemental Fig. 5F).

BTG2 temporally restricts AngII-induced activationkinetics

We decided to investigate how BTG2 controls the AngIIgene expression response. BTG2 promotes generalRNA decay through deadenylation (Mauxion et al. 2008)and represses aldosterone production (Fig. 5B). BothBTG2 mRNA (Fig. 6A) and protein (Fig. 6B) were rapidlyinduced by AngII stimulation each peaking at ∼3 h.Therefore, we hypothesize that BTG2 actively promotesthe resolution of aldosterone production. We performedRNA-seq of siRNA knockdown of BTG2 in H295R cells

B

A

C

D

FIGURE 5. Regulation of aldosterone by RBPs in AngII stimulatedcells. (A) Boxplots of all RBPs (gray) and candidate RBPs (red) forchange in expression upon AngII stimulation (left), human adrenal tis-sue expression levels (center), and adrenal tissue specificity scores(right). Bar plot of the change in aldosterone concentrations for knock-downs resulting in (B) increased aldosterone concentration, (C ) de-creased aldosterone concentration, and (D) discordant or singlesiRNA results. Aldosterone levels weremeasured using ELISA from su-pernatants of H295R cells electroporatedwith siRNAs targeting candi-date RBPs and stimulatedwith AngII for 24 h. The y-axis represents thechange in aldosterone concentration versus mock electroporation(seeMaterials andMethods). The error bars represent the standard er-ror of at least six replicates. RBPs are color-coded by their knownfunction.

Fu et al.

940 RNA (2021) Vol. 27, No. 8

that were unstimulated or AngII-stimulated for 6 h and 24h and performed RNA-seq in duplicate (Supplemental Ta-ble 9). As expected, BTG2 mRNA levels were lower acrossall time points in the BTG2 depletion compared to themock depletion (Supplemental Fig. 6A). Next, we as-sessed the overlap between mRNAs up-regulated ordown-regulated by BTG2 depletion with the AngII-re-sponse clusters. We found that mRNAs up-regulatedupon BTG2 depletion were strongly enriched in the as3,as4 classes and modestly enriched in aw1–aw4 classes(Fig. 6C). This is consistent with the propensity for theseclusters to contain mRNAs exhibiting increases in RNAdecay during steroidogenesis (Fig. 3C). Transcripts en-coding critical prosteroidogenic factors and enzymesthat were induced by AngII exhibited increased up-regu-lation in the BTG2 depleted cells (Fig. 6D), whileunchanged mRNAs were not altered by BTG2 (Fig. 6D,right). Altogether, these data are consistent with a modelin which BTG2 is induced by AngII to up-regulate the de-cay of mRNAs encoding prosteroidogenic factors to pre-vent overproduction of aldosterone.

DISCUSSION

Despite our current understanding ofadrenocortical steroidogenesis, thefundamental regulatory networksthat govern aldosterone homeostasisremain unclear. Our data indicatethat AngII stimulation increases bothmRNAproduction and decay to rapid-ly implement and resolve a prosteroi-dogenic gene expression program.The AngII-induced expression dy-namics in H295R cells are likely to bephysiologically relevant since theywere largely recapitulated in primaryhuman adrenocortical cells stimulatedwith AngII and ACTH. This regula-tory scheme enables the robustproduction of aldosterone whilealso preventing overproduction. Thedepletion of multiple factors promot-ing RNA decay, which themselveswere up-regulated by AngII, resultedin an excess of aldosterone. The addi-tional energy expenditure of increas-ing both transcription and decaymay represent the cellular cost forthe absence of a mechanism to storealdosterone within the cell and re-lease it in response to stimulus. It willbe interesting to determine if this cou-pling is disrupted by mutations driv-ing primary aldosteronism.

RNA decay temporally coordinates aldosteroneproduction

The stimulation response of adrenocortical cells to AngIIresembles an impulse (or single pulse) pattern consistentwith functional higher order temporal coordination. Thissystem exhibits both regulatory modules (12 clusters inFig. 1D), which consists of coexpressed genes with thesame temporal pattern, which reflects a particular sequen-tial order or cascade of gene or module expression (Yosefand Regev 2011). The mRNAs in the regulatory modulesencoded functionally related proteins that play distinctroles at specific times during the AngII response. For ex-ample, expression levels of mRNAs encoding transcriptionfactors and RBPs peaked within the first few hours, whilethose encoding steroidogenic enzymes peaked 6–8 hpost stimulation. Indeed, the temporal coordination of ex-pression of these distinct factors is consistent with previouscharacterization of early and late steps of steroidogenesis(Nogueira et al. 2009). Our data suggest that both tran-scriptional cascades and intrinsic differences in RNA decay

E

B

A

C

D

FIGURE 6. BTG2 constrains AngII RNA expression kinetics to prevent aldosterone overpro-duction. (A) Line plot of fold change in BTG2 mRNA levels in response to AngII. (B) Westernblot of BTG2 and B-Tubulin protein levels in response to AngII. (C ) Heatmap of the odds ratiofor the overlap between response cluster membership and RNAs exhibiting down-regulation,no change, or up-regulation upon BTG2 siRNA depletion 24 h after AngII stimulation (yellowindicates more overlap). (D) Bar plot of expression levels of mRNAs encoding key AngII-in-duced steroidogenic proteins exhibiting statistically significant up-regulation (P<0.05,DESeq2) upon BTG2 siRNA depletion during AngII stimulation. Error bars represent standarddeviation. (E) Model for incoherent feedforward loop using increases in transcription and RNAdecay to coordinate AngII stimulation kinetics to ensure the appropriate production ofaldosterone.

RNA regulatory control of aldosterone production

www.rnajournal.org 941

rates determine the kinetics of the AngII gene expressionresponse. Specifically, the decay rate of transcriptionallyinduced mRNAs is inversely proportional to the time topeak expression. Therefore, RNA decay is critical for tem-porally coordinating the response time of these regulatorymodules and consequently aldosterone homeostasis. Ourfindings are similar to the observed role RNA decay in co-ordinating immune and inflammation responses (Hao andBaltimore 2009; Elkon et al. 2010; Rabani et al. 2011). Thetemporal coordination of these regulatory modules bypost-transcriptional regulation may be examples of dy-namic RNA regulons (Keene 2007). Interestingly, these ob-servations occur in systems that require a robust, rapid, yetmeasured production of a physiologically potent moleculein either cytokines or steroid hormones.

Coupling of transcription and decay dynamicscontrols aldosterone homeostasis

Multiple lines of evidence indicate that one or more mixedincoherent feedforward loops (MIFFL) ensure the properpromotion and resolution of steroidogenesis (Fig. 6E).Specifically, AngII promotes the transcriptional up-regula-tion of both prosteroidogenic factors (STAR, CYP11B2)and RNA metabolism factors (BTG2, ZFP36L2) that nega-tively regulate the expression of prosteroidogenic factors.These Incoherent feedforward loops have largely been ex-amined with respect to transcription factors and miRNAsand are a recurrent motif to enhance the robustness(Shalgi et al. 2007; Tsang et al. 2007). Transcription factorand RBP versions of these motifs remain poorly character-ized (Joshi et al. 2012), with the exception of the RBPZFP36 (TTP) in macrophage activation models (Rabaniet al. 2014). Nevertheless, the behavioral benefits associat-ed with incoherent feedforward loops include the produc-tion of pulse-like patterns, faster response times tostimulation, output noise buffering, and even detectionof fold-change in expression (Basu et al. 2004; Manganet al. 2006; Goentoro et al. 2009; Siciliano et al. 2013).These features would be invaluable for facilitating the rap-id production and resolution of the response to AngII stim-ulation to maintain proper aldosterone homeostasis.However, due to a combination of technical limitationsand need for additional experimentation, we do notknow which of those features that BTG2 contributes to.H295R cells are difficult to transfect and it is imperativeto develop stable models for controlling gene dose, whichwe are developing using CRISPR/Cas9. Additionally, wewould gain sensitivity by using recently developed meth-ods that provide a direct readout of newly synthesized ver-sus preexisting RNAs in the same sample (Herzog et al.2017; Schofield et al. 2018). Thus, we are likely underesti-mating the true number of RNAs exhibiting changes in de-cay rates during steroidogenesis. Furthermore, we do notknow the identity of the transcription factor(s) promoting

synthesis of mRNAs encoding prosteroidogenic proteinsand BTG2, though SF1 (NR5A1) and CREB1 are the mostlikely candidates (Caron et al. 1997; Clark and Combs1999; Nogueira and Rainey 2010; Selvaraj et al. 2018).The RBPs ZFP36L1 and ZFP36L2 are both induced byAngII and bind to AREs in the 3′-UTR of target mRNA topromote decay and may be additional repressors.Consistent with our expectation, ZFP36L1 was shown tosuppress ACTH-stimulated STAR mRNA induction in bo-vine adrenocortical cells and mouse cell lines (Duan et al.2009). However, this was mediated via ZFP36L1 bindingto AREs in an alternative 3′-UTR that we did not detect inour RNA-seq data from H295R or human adrenocorticalcells. Finally, BTG2 depletion only deregulated a subsetof mRNAs, which suggests that BTG2 may synergize withspecific RNA-decay pathways as previously postulated(Stupfler et al. 2016).

Dynamic RNA regulatory network controllingsteroidogenesis

We have uncovered a novel paradigm by which RBAs andregulated RNA decay control the kinetics of AngII-stimu-lated gene expression to facilitate the implementationand resolution of prosteroidogenic gene expression andmetabolic programs required for proper aldosterone pro-duction in humans. In addition to RNA decay factors, weidentified the RNA modification enzyme TRUB1 was alsoa repressor of steroidogenesis. TRUB1 is a pseudouridinesynthase that modifies tRNAs and mRNAs (Safra et al.2017). We also identified numerous activators of aldoste-rone production. We found that the signaling scaffold pro-tein AKAP1 activates aldosterone production, which isconsistent with a proposed role for AKAP1 in promotingthe localized translation of STAR mRNA at the mitochon-dria (Dyson et al. 2008; Grozdanov and Stocco 2012).Additional activators identified include the cytoplasmicRBPs CPEB4, MSI2, and PELO, which regulated stabilityand translation, as well as, MBNL2 a well-known splicingfactor. The majority of these RBPs have not previouslybeen associated with aldosterone production or steroido-genesis. The identification of activators and repressors isnot surprising given the dysregulation of aldosterone ho-meostasis has pathological consequences. Dissecting therole of combinatorial regulation by RBPs in the dynamicsteroidogenic regulatory network will undoubtedly revealnovel points of control. Since the adrenal cortex is amena-ble to the delivery of modified oligonucleotides (Biscanset al. 2019), the RBP–RNA regulatory interactions may betargeted for disruption to precisely and specifically modu-late steroidogenesis.

MATERIALS AND METHODS

See Supplemental Methods for more details.

Fu et al.

942 RNA (2021) Vol. 27, No. 8

Cell culture

H295R cells were cultured in complete media containing DMEM:F12 with 10% Cosmic Calf Serum (Hyclone: SH30087.03) and 1%ITS+ Premix (Corning: 354352). H295R cells were stimulated with10 nM Angiontensin II (Sigma: A9525) after being in low sera me-dia for 24 h (DMEM/F12 with 0.1% CCS, 1% ITS).

Primary human adrenal cells were isolated and cultured as de-scribed previously (Xing et al. 2010; Rege et al. 2015). Primary ad-renal cells were plated at a density of 20,000 cells/well (48 welldish) in growth medium and grown to 60% confluence after whichthey were starved in low serum medium 18 h prior to treatmentwith either 10 nM AngII or 10 µM ACTH.

Gene expression measurements

H295R cells were collected in TRIzol and RNA was isolated usingZYMO Research DirectZol Miniprep Plus kit following the manu-facturer’s instructions with on-column DNase I digestion. RNAwas collected from ex vivo stimulated primary adrenal cells usingtheQiagen RNEasyMiniPrep Plus kit following themanufacturer’sinstructions and eluted into nuclease-free water.

RT-qPCR

Reverse transcription was performed using 100 ng of RNA inputfor the Bio-Rad iScripit kit and qPCR using Bio-Rad iTaqUniversal SYBR Green Supermix on a Bio-Rad CFX 384 qPCR in-strument. Analysis of qPCR was performed using the delta-deltaCq method normalizing to GAPDH and unstimulated controls.

Metabolic labeling experiments

We performed 4-thiouridine (4sU) experiments as described pre-viously (Mukherjee et al. 2017). Briefly, H295R cells were treatedfor a 20-min pulse with 500 µM 4sU. RNA was extracted and splitinto input and biotinylated samples. The biotinylated RNA wasimmunoprecipitated using streptavidin beads and both inputand immunoprecipitated RNAwere subject to rRNA depletion us-ing RiboZero and sequenced using the qRNA-seq kit (BIOOscientific).

RNA-seq and regulatory analysis

Salmon (Patroet al. 2017)wasused for quantifying transcript levelsfrom all libraries using Gencode v26. A custom salmon index con-taining a Gencode v26 precursor, mature transcripts, and ERCCspike-ins was used as previously described (Mukherjee et al.2017). All downstream analysis was performed in R and detailedin the Supplemental Methods. Unstimulated (initial) RNA decayrates were estimated using 4sU metabolic labeling data and theINSPEcT R package (de Pretis et al. 2015). Overrepresentation of7mers was calculated via Cwords (Rasmussen et al. 2013) from3′UTR sequences ranked by unstimulated RNA decay rates. RBPmotifs enrichment was calculated using https://github.com/TaliaferroLab/FeatureReachR. Determining RNA stability changesduring the time course used salmon-quantified intronic and exon-

ic reads, adapting the Exon-Intron Split Analysis concept imple-mented in eisaR (Gaidatzis et al. 2015).

Aldosterone screen

H295R cells (2.5M) cultured in complete media were electropo-rated with 10 µM siRNA (Thermo Fisher Scientific—see supple-ment for catalog numbers) and plated in a 96-well plate. After24 h, half the wells hadmedia that was replaced with low serame-dia and the other half with low sera media containing AngII. After24 h, aldosterone levels in the supernatant were measured usingAldosterone Competitive ELISA kit (ThermoFisher) and cell viabil-ity was measured from the cells using PrestoBlue reagent (Ther-moFisher) by reading fluorescence on a BioTek Synergy HTplate reader. Each plate contained a mock transfection that wenormalized to for calculating aldosterone levels in each siRNA ex-periment while taking plate batch effects and cell viability intoaccount.

DATA DEPOSITION

Raw sequencing data have been deposited in SRA GSE163801.Code and processed data are available at https://github.com/mukherjeelab/2021_RNA_PTR-steroid-kinetics.

SUPPLEMENTAL MATERIAL

Supplemental material is available for this article.

ACKNOWLEDGMENTS

Wewould like to thankDavid Bentley and Lori Sussel for their sup-port, collegiality, and critical review of the manuscript, as well asUwe Ohler and Austen Gillen for initial support for this study. Wethank Bertrand Seraphim for anti-BTG2 antibody. This work wassupported by the American Heart Association Predoctoral Fellow-ship Award 20PRE35220016 (K.W.), the Predoctoral TrainingGrant in Molecular Biology NIH-T32-GM008730 (K.W.), the Uni-versity of Colorado Anschutz Medical Campus RNA BioscienceInitiative (N.M., R.F.), Boettcher Foundation Webb-Waring EarlyCareer Investigator Award AWD-103075 (N.M.), and National In-stitutes of Health DK043140 (W.R.).Author contributions: N.M. conceived the project; K.Y.W., A.B.,

A.H., K.W., and J.R. performed experiments and collected data;R.F, K.Y.W., K.R., and N.M. performed formal analysis and con-ducted the visualization; N.M. wrote the original draft; R.F., K.Y.W., K.R., W.R, and N.M. reviewed and edited the paper; K.Y.W.,W.R., and N.M. acquired funding; W.R. and N.M. provided re-sources; N.M. supervised the project.

Received February 24, 2021; accepted May 20, 2021.

REFERENCES

Albert TK, Lemaire M, van Berkum NL, Gentz R, Collart MA,Timmers HT. 2000. Isolation and characterization of human ortho-logs of yeast CCR4-NOT complex subunits. Nucleic Acids Res 28:809–817. doi:10.1093/nar/28.3.809

RNA regulatory control of aldosterone production

www.rnajournal.org 943

Alkallas R, Fish L, Goodarzi H, Najafabadi HS. 2017. Inference of RNAdecay rate from transcriptional profiling highlights the regulatoryprograms of Alzheimer’s disease. Nat Commun 8: 909. doi:10.1038/s41467-017-00867-z

Bassett MH, Suzuki T, Sasano H, White PC, Rainey WE. 2004. The or-phan nuclear receptors NURR1 and NGFIB regulate adrenal aldo-sterone production.Mol Endocrinol 18: 279–290. doi:10.1210/me.2003-0005

Basu S, Mehreja R, Thiberge S, Chen M-T, Weiss R. 2004.Spatiotemporal control of gene expression with pulse-generatingnetworks. Proc Natl Acad Sci 101: 6355–6360. doi:10.1073/pnas.0307571101

Biscans A, Coles A, Haraszti R, Echeverria D, Hassler M, Osborn M,Khvorova A. 2019. Diverse lipid conjugates for functional extra-he-patic siRNA delivery in vivo. Nucleic Acids Res 47: 1082–1096.doi:10.1093/nar/gky1239

Brown JM, Siddiqui M, Calhoun DA, Carey RM, Hopkins PN,Williams GH, Vaidya A. 2020. The unrecognized prevalence of pri-mary aldosteronism: a cross-sectional study. Ann Intern Med 173:10–20. doi:10.7326/M20-0065

Caron KM, Ikeda Y, Soo SC, Stocco DM, Parker KL, Clark BJ. 1997.Characterization of the promoter region of the mouse gene en-coding the steroidogenic acute regulatory protein. MolEndocrinol 11: 138–147. doi:10.1210/mend.11.2.9880

Clark BJ, Combs R. 1999. Angiotensin II and cyclic adenosine 3′,5′-monophosphate induce human steroidogenic acute regulatoryprotein transcription through a common steroidogenic factor-1 el-ement. Endocrinology 140: 4390–4398. doi:10.1210/endo.140.10.7085

de Pretis S, Kress T, Morelli MJ, Melloni GEM, Riva L, Amati B,Pelizzola M. 2015. INSPEcT: a computational tool to infer mRNAsynthesis, processing and degradation dynamics from RNA- and4sU-seq time course experiments. Bioinformatics 31: 2829–2835. doi:10.1093/bioinformatics/btv288

Duan H, Cherradi N, Feige J-J, Jefcoate C. 2009. cAMP-dependentposttranscriptional regulation of steroidogenic acute regulatory(STAR) protein by the zinc finger protein ZFP36L1/TIS11b. MolEndocrinol 23: 497–509. doi:10.1210/me.2008-0296

Dyson MT, Jones JK, Kowalewski MP, Manna PR, Alonso M,GottesmanME, Stocco DM. 2008.Mitochondrial A-kinase anchor-ing protein 121 binds type II protein kinase A and enhances ste-roidogenic acute regulatory protein-mediated steroidogenesis inMA-10 mouse leydig tumor cells. Biol Reprod 78: 267–277.doi:10.1095/biolreprod.107.064238

Elkon R, Zlotorynski E, Zeller KI, Agami R. 2010. Major role for mRNAstability in shaping the kinetics of gene induction. BMCGenomics11: 259. doi:10.1186/1471-2164-11-259

Feldman RD. 2014. Aldosterone and blood pressure regulation: re-cent milestones on the long and winding road from electrocortinto KCNJ5, GPER, and beyond. Hypertension 63: 19–21. doi:10.1161/HYPERTENSIONAHA.113.01251

Gaidatzis D, Burger L, Florescu M, Stadler MB. 2015. Analysis ofintronic and exonic reads in RNA-seq data characterizes transcrip-tional and post-transcriptional regulation.Nat Biotechnol 33: 722–729. doi:10.1038/nbt.3269

Gerstberger S, Hafner M, Tuschl T. 2014. A census of human RNA-binding proteins. Nat Rev Genet 15: 829–845. doi:10.1038/nrg3813

Goentoro L, Shoval O, Kirschner MW, Alon U. 2009. The incoherentfeedforward loop can provide fold-change detection in gene reg-ulation. Mol Cell 36: 894–899. doi:10.1016/j.molcel.2009.11.018

Gross F, Gysel H. 1954. The action of electrocortin in the adrenalecto-mized dog. Eur J Endocrinol 15: 199–209. doi:10.1530/acta.0.0150199

Grozdanov PN, Stocco DM. 2012. Short RNA molecules with highbinding affinity to the KH motif of A-kinase anchoring protein 1(AKAP1): implications for the regulation of steroidogenesis. MolEndocrinol 26: 2104–2117. doi:10.1210/me.2012-1123

Hao S, Baltimore D. 2009. The stability of mRNA influences the tem-poral order of the induction of genes encoding inflammatory mol-ecules. Nat Immunol 10: 281–288. doi:10.1038/ni.1699

Hattangady NG, Olala LO, Bollag WB, Rainey WE. 2012. Acute andchronic regulation of aldosterone production. Mol CellEndocrinol 350: 151–162. doi:10.1016/j.mce.2011.07.034

Herzog VA, Reichholf B, Neumann T, Rescheneder P, Bhat P,Burkard TR, Wlotzka W, von Haeseler A, Zuber J, Ameres SL.2017. Thiol-linked alkylation of RNA to assess expression dynam-ics. Nat Methods 14: 1198–1204. doi:10.1038/nmeth.4435

Hui J, ReitherG, Bindereif A. 2003.Novel functional role of CA repeatsand hnRNP L in RNA stability. RNA 9: 931–936. doi:10.1261/rna.5660803

Joshi A, Beck Y, Michoel T. 2012. Post-transcriptional regulatory net-works play a key role in noise reduction that is conserved from mi-cro-organisms tomammals. FEBS J 279: 3501–3512. doi:10.1111/j.1742-4658.2012.08571.x

Keene JD. 2007. RNA regulons: coordination of post-transcriptionalevents. Nat Rev Genet 8: 533–543. doi:10.1038/nrg2111

Lebrethon MC, Naville D, Begeot M, Saez JM. 1994. Regulation ofcorticotropin receptor number andmessenger RNA in cultured hu-man adrenocortical cells by corticotropin and angiotensin II. J ClinInvest 93: 1828–1833. doi:10.1172/JCI117168

Lin D, Sugawara T, Strauss JF III, Clark BJ, Stocco DM, Saenger P,Rogol A, Miller WL. 1995. Role of steroidogenic acute regulatoryprotein in adrenal and gonadal steroidogenesis. Science 267:1828–1831. doi:10.1126/science.7892608

Luscombe NM, Babu MM, Yu H, Snyder M, Teichmann SA,Gerstein M. 2004. Genomic analysis of regulatory network dynam-ics reveals large topological changes. Nature 431: 308–312.doi:10.1038/nature02782

Mangan S, Itzkovitz S, Zaslaver A, Alon U. 2006. The incoherent feed-forward loop accelerates the response-time of the gal system ofEscherichia coli. J Mol Biol 356: 1073–1081. doi:10.1016/j.jmb.2005.12.003

Masumura Y, Higo S, Asano Y, Kato H, Yan Y, Ishino S, Tsukamoto O,Kioka H, Hayashi T, Shintani Y, et al. 2016. Btg2 is a negative reg-ulator of cardiomyocyte hypertrophy through a decrease in cyto-solic RNA. Sci Rep 6: 28592. doi:10.1038/srep28592

Mauxion F, Faux C, Séraphin B. 2008. The BTG2 protein is a generalactivator of mRNA deadenylation. EMBO J 27: 1039–1048. doi:10.1038/emboj.2008.43

Milek M, Imami K, Mukherjee N, Bortoli FD, Zinnall U, Hazapis O,Trahan C, Oeffinger M, Heyd F, Ohler U, et al. 2017. DDX54 reg-ulates transcriptome dynamics during DNA damage response.Genome Res 27: 1344–1359. doi:10.1101/gr.218438.116

Monticone S, D’Ascenzo F, Moretti C, Williams TA, Veglio F, Gaita F,Mulatero P. 2018. Cardiovascular events and target organ damagein primary aldosteronism compared with essential hypertension: asystematic review and meta-analysis. Lancet Diabetes Endocrinol6: 41–50. doi:10.1016/S2213-8587(17)30319-4

Mukherjee N, Calviello L, Hirsekorn A, de Pretis S, Pelizzola M,Ohler U. 2017. Integrative classification of human coding and non-coding genes through RNA metabolism profiles. Nat Struct MolBiol 24: 86–96. doi:10.1038/nsmb.3325

Mukherjee N, Wessels H-H, Lebedeva S, Sajek M, Ghanbari M,Garzia A, Munteanu A, Yusuf D, Farazi T, Hoell JI, et al. 2019.Deciphering human ribonucleoprotein regulatory networks.Nucleic Acids Res 47: 570–581. doi:10.1093/nar/gky1185

NanbaK,OmataK,Gomez-SanchezCE, StratakisCA,DemidowichAP,Suzuki M, Thompson LDR, Cohen DL, Luther JM, Gellert L, et al.

Fu et al.

944 RNA (2021) Vol. 27, No. 8

2019. Genetic characteristics of aldosterone-producing adenomasin blacks. Hypertension 73: 885–892. doi:10.1161/HYPERTENSIONAHA.118.12070

Nogueira EF, RaineyWE. 2010. Regulation of aldosterone synthase byactivator transcription factor/cAMP response element-bindingprotein family members. Endocrinology 151: 1060–1070. doi:10.1210/en.2009-0977

Nogueira EF, Bollag WB, Rainey WE. 2009. Angiotensin II regulationof adrenocortical gene transcription. Mol Cell Endocrinol 302:230–236. doi:10.1016/j.mce.2008.08.024

Palumbo MC, Farina L, Paci P. 2015. Kinetics effects and modeling ofmRNA turnover. Wiley Interdiscip Rev RNA 6: 327–336. doi:10.1002/wrna.1277

Parmar J, Key RE, Rainey WE. 2008. Development of an adrenocorti-cotropin-responsive human adrenocortical carcinoma cell line. JClin Endocrinol Metab 93: 4542–4546. doi:10.1210/jc.2008-0903

Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. 2017. Salmonprovides fast and bias-aware quantification of transcript expres-sion. Nat Methods 14: 417–419. doi:10.1038/nmeth.4197

Rabani M, Levin JZ, Fan L, Adiconis X, Raychowdhury R, Garber M,Gnirke A, Nusbaum C, Hacohen N, Friedman N, et al. 2011.Metabolic labeling of RNA uncovers principles of RNA productionand degradation dynamics in mammalian cells.Nat Biotechnol 29:436–442. doi:10.1038/nbt.1861

Rabani M, Raychowdhury R, Jovanovic M, Rooney M, Stumpo DJ,Pauli A, Hacohen N, Schier AF, Blackshear PJ, Friedman N, et al.2014. High-resolution sequencing andmodeling identifies distinctdynamic RNA regulatory strategies. Cell 159: 1698–1710. doi:10.1016/j.cell.2014.11.015

Rasmussen SH, Jacobsen A, Krogh A. 2013. cWords: systematicmicroRNA regulatory motif discovery from mRNA expressiondata. Silence 4: 2. doi:10.1186/1758-907X-4-2

Rege J, Nishimoto HK, Nishimoto K, Rodgers RJ, Auchus RJ,Rainey WE. 2015. Bone morphogenetic protein-4 (BMP4): a para-crine regulator of human adrenal C19 steroid synthesis.Endocrinology 156: 2530–2540. doi:10.1210/en.2014-1942

Romero DG, Plonczynski M, Vergara GR, Gomez-Sanchez EP, Gomez-Sanchez CE. 2004. Angiotensin II early regulated genes in H295Rhuman adrenocortical cells. Physiol Genomics 19: 106–116.doi:10.1152/physiolgenomics.00097.2004

Romero DG, Rilli S, Plonczynski MW, Yanes LL, Zhou MY, Gomez-Sanchez EP, Gomez-Sanchez CE. 2007. Adrenal transcriptionregulatory genes modulated by angiotensin II and their role insteroidogenesis. Physiol Genomics 30: 26–34. doi:10.1152/physiolgenomics.00187.2006

SafraM, Nir R, Farouq D, Vainberg Slutskin I, Schwartz S. 2017. TRUB1is the predominant pseudouridine synthase acting on mammalianmRNA via a predictable and conserved code. Genome Res 27:393–406. doi:10.1101/gr.207613.116

Schofield JA, Duffy EE, Kiefer L, Sullivan MC, Simon MD. 2018.TimeLapse-seq: adding a temporal dimension to RNA sequencingthrough nucleoside recoding. Nat Methods 15: 221–225. doi:10.1038/nmeth.4582

Selvaraj V, Stocco DM, Clark BJ. 2018. Current knowledge on theacute regulation of steroidogenesis. Biol Reprod 99: 13–26.doi:10.1093/biolre/ioy102

ShalemO, DahanO, LevoM,MartinezMR, Furman I, Segal E, Pilpel Y.2008. Transient transcriptional responses to stress are generatedby opposing effects of mRNA production and degradation. MolSyst Biol 4: 223. doi:10.1038/msb.2008.59

Shalgi R, Lieber D, Oren M, Pilpel Y. 2007. Global and local architec-ture of the mammalian microRNA-transcription factor regulatorynetwork. PLoS Comput Biol 3: e131. doi:10.1371/journal.pcbi.0030131

Shim J, Lim H, Yates J R, Karin M. 2002. Nuclear export of NF90 is re-quired for interleukin-2 mRNA stabilization. Mol Cell 10: 1331–1344. doi:10.1016/S1097-2765(02)00730-X

Siciliano V, Garzilli I, Fracassi C, Criscuolo S, Ventre S, di Bernardo D.2013. MiRNAs confer phenotypic robustness to gene networks bysuppressing biological noise. Nat Commun 4: 2364. doi:10.1038/ncomms3364

Smale ST. 2012. Transcriptional regulation in the innate immunesystem. Curr Opin Immunol 24: 51–57. doi:10.1016/j.coi.2011.12.008

Stupfler B, Birck C, Séraphin B, Mauxion F. 2016. BTG2 bridgesPABPC1 RNA-binding domains and CAF1 deadenylase tocontrol cell proliferation. Nat Commun 7: 10811. doi:10.1038/ncomms10811

Tamayo P, Mootha VK, Mukherjee S. 2005. Gene set enrichment anal-ysis: a knowledge-based approach for interpreting genome-wideexpression profiles. Proc Natl Acad Sci 102: 15545–15550.doi:10.1073/pnas.0506580102

Tsang J, Zhu J, vanOudenaarden A. 2007.MicroRNA-mediated feed-back and feedforward loops are recurrent network motifs in mam-mals. Mol Cell 26: 753–767. doi:10.1016/j.molcel.2007.05.018

Wang T, Rowland JG, Parmar J, Nesterova M, Seki T, Rainey WE.2012. Comparison of aldosterone production among human adre-nocortical cell lines.HormMetab Res 44: 245–250. doi:10.1055/s-0031-1298019

Xing Y, Parker CR, Edwards M, Rainey WE. 2010. ACTH is a potentregulator of gene expression in human adrenal cells. J MolEndocrinol 45: 59–68. doi:10.1677/JME-10-0006

Xing Y, Edwards MA, Ahlem C, Kennedy M, Cohen A, Gomez-Sanchez CE, Rainey WE. 2011. The effects of ACTH on steroidmetabolomic profiles in human adrenal cells. J Endocrinol 209:327–335. doi:10.1530/JOE-10-0493

Yosef N, Regev A. 2011. Impulse control: temporal dynamics in genetranscription. Cell 144: 886–896. doi:10.1016/j.cell.2011.02.015

RNA regulatory control of aldosterone production

www.rnajournal.org 945